Maan, M.E.
Citation
Maan, M. E. (2006, May 11). Sexual selection and speciation: mechanisms in Lake Victoria
cichlid fish. Retrieved from https://hdl.handle.net/1887/4382
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in theInstitutional Repository of the University of Leiden Downloaded from: https://hdl.handle.net/1887/4382
Chapter 2
‘It is shown by various facts, given hereafter, and by the
results fairly attributable to sexual selection, that the
female, though comparatively passive, generally exerts
some choice and accepts one male in preference to others.
Or she may accept, as appearances would sometimes lead
us to believe, not the male which is the most attractive to
Intraspecific sexual selection on a speciation trait,
male coloration, in the Lake Victoria cichlid
Pundamilia nyererei
Martine E. Maan, Ole Seehausen, Linda Söderberg, Lisa Johnson, Erwin A. P. Ripmeester, Hillary D. J. Mrosso, Martin I. Taylor, Tom J.M. Van Dooren and Jacques J.M. van Alphen
Introduction
The role of selection in the origin of new species is a relatively poorly understood problem in evolutionary biology and theoretical studies amply outnumber ex-perimental ones (Kirkpatrick & Ravigne 2002). However, without empirical data on natural populations, it is impossible to judge how realistic model assumptions are.
The cichlid fish species flocks of the Great Lakes of East Africa, and the haplochromine radiations in Lakes Victoria and Malawi in particular, are classical examples of rapid speciation and adaptive radiation within the confines of single lakes. Lake Victoria is the youngest of the African Great Lakes (0.25-0.75 My; Fryer 1996) and sedimentological and paleoclimatic evidence indicates that it was completely dry as recently as 14.500 years ago (Johnson et al. 1996; Johnson et al. 2000). Although the genetic diversity contained in the flock must be far older (Nagl et al. 2000), most of the 500 or more endemic species must have diverged in an extremely short time. Several different models of speciation involving selection on mate choice traits have been proposed to explain this (Kocher 2004). Theoreti-cally, mate choice can establish reproductive isolation very rapidly, since it directly influences mating behaviour and does not require postzygotic reinforcement through ecological selection against hybrids (Kirkpatrick & Ravigne 2002). Em-pirical support for sexual selection as a diversifying force has been demonstrated in birds (Barraclough et al. 1995; Uy & Borgia 2000; Irwin et al. 2001), lizards (Stuart-Fox & Owens 2003), parrotfishes (Streelman et al. 2002), insects (Arnqvist et al. 2000; Gray & Cade 2000), spiders (Masta & Maddison 2002), and snails (Schilthuizen 2003) but the evidence is not always unambiguous (e.g. Boake et al. 1997; Houde & Hankes 1997; Gage et al. 2002).
D I R E C T I O N A L S E X U A L S E L E C T I O N Pundamilia pundamilia (Seehausen et al. 1998a) and Pundamilia nyererei (Witte-Maas & Witte 1985) are rock-dwelling haplochromine cichlids endemic to Lake Victoria.
P. nyererei is sympatric with P. pundamilia in all of its range, the two are very closely related and exchange genes in some localities (Seehausen 1996; Taylor et al. in prep). They differ only slightly in anatomy and both species exhibit vertical mela-nin bars that are more pronounced in males than in females. However, whereas females of both species are cryptically coloured and can be distinguished only with difficulty, P. pundamilia males are metallic blue-grey and P. nyererei males are bright red dorsally and yellow laterally. This difference in male colour pattern is important in female mate choice: in laboratory experiments, females choose con-specific males when colours are visible, but not when they are masked (Seehausen & Van Alphen 1998).
Here we test within P. nyererei whether females select for brighter red males, using three approaches. First, we conduct mate choice experiments in the laboratory to test whether females prefer the more red one of two males, and whether the strength of preference increases with the magnitude of the difference in redness. Second, we investigate the importance of male coloration as a predic-tor of female responsiveness in nature, where female choice could also be influ-enced by extended phenotype traits and other sources of variation. Third, we test whether male coloration predicts male mating success in a mesocosm experiment.
Methods
Fish
We studied a P. nyererei population at Makobe Island in the western Speke Gulf (Tanzania; Seehausen & Bouton 1997), where the water is relatively clear (Secchi reading mean±se=221±15 cm in the study period). All experiments and observa-tions were carried out in 2001. Mature P. nyererei males defend territories on rocky bottom at four to seven meters water depth. They attract females by vigor-ous courtship displays that resemble those of other haplochromine species (See-hausen & Van Alphen 1998). Courtship typically starts with Lateral Display, in which the male positions itself perpendicular to the female and spreads all fins. This is followed by Quiver, a high-frequency shaking movement of the body. Fi-nally, the male leads the female to the centre of the territory in a quick swimming bout, often with exaggerated tail beats (Lead Swim). Mating occurs in rocky crev-ices, and immediately after spawning the mouthbrooding female leaves the terri-tory. Females mouthbrood eggs and larvae for about three weeks, and guard the fry for an additional week after releasing them (Seehausen 1996).
Mate choice experiment in the laboratory
enclo-sure, shelter was provided by a brick on top of a PVC tube. These shelters were readily accepted by the males as the centres of their territories. In the middle of the main tank two large stones provided shelter for the test female and ensured that the males could not see each other. The water in the male enclosures was fil-tered internally, making chemical communication impossible. Air stones were pre-sent in both enclosures and in the main tank; water temperature was kept at 24-26°C.
We used 25 wild caught P. nyererei: 14 males, assembled into 7 male pairs of fixed composition, and 11 females. All males were photographed, measured and weighed. Pairs were assembled such that the redscore differences between males in a pair varied from 1 to 30% body coverage. Size differences between paired males were small and did not co-vary with colour differences (standard length (SL): high redscore: 95.3±5.7 mm, low redscore: 95.9±4.5 mm [paired t-test:
t6=-0.54, p=0.61]; weight: high redscore: 27.3±7.3 g, low redscore: 27.0±3.0 g
[paired t-test: t6=0.17, p=0.87]).
Males were released into their compartments the evening before trials. A maximum of three females was tested with one male pair in a day and at least two hours separated consecutive trials. Male pairs were exchanged after three success-ful trials or after two days. The position of the males (left or right side of the aquarium) was reversed each time the same pair was introduced. To start a trial, the test female was released in the middle of the tank. Behaviour was recorded with Observer 3.0 software (Noldus). Observation time started when the test fe-male was within 30 cm of either one of the fe-male compartments and stopped when she left this area. This distance was chosen because, in field observations, any fish within 30 cm of a P. nyererei territory elicits a behavioural response of the territory owner, indicating 30 cm as an interaction threshold distance. Trials were com-pleted once 15 minutes of observation time had been collected.
Each female was tested once with every male pair, but in the analysis we in-cluded only those trials in which both males courted (performing Quiver or Lead Swim at least once) and the test female responded positively (approaching a male in response to his Lateral Display, Quiver or Lead Swim) to at least one of them (48 trials). For every male courtship display event, we recorded whether the fe-male responded by approaching the fe-male. The resulting proportion is our meas-ure of female response to each male. The difference in female response to two males in a trial is our measure of female preference. We also recorded aggressive behaviour of the males (butting and biting attempts) and counted the total num-ber of encounters of each male with the test female. Courtship intensity was de-fined as the mean number of courtship behaviours per encounter with the female and was calculated separately for Lateral Display, Quiver and Lead Swim.
Field observations
min-D I R E C T I O N A L S E X U A L S E L E C T I O N
utes duration each for each male, yielding a total of between 50 and 100 minutes observation time per male (total observation time: 1950 minutes). Observations were carried out between 09:00 and 13:00 hours and all observations for one male were completed within a period of 6 to 32 days. We recorded all interactions be-tween these males and conspecific females. Female response was defined as the proportion of courted females responding positively (i.e. approaching the male) to male courtship. Aggressive behaviour towards females, including aggressive Lat-eral Display (interpreted as such when it was followed by Frontal Display or chas-ing or bitchas-ing), was much more common in these observations than in the labora-tory experiments. We therefore used the number of times a male quivered to a female as our measure of male courtship intensity. Quiver is only rarely displayed in antagonistic contexts, unless the antagonists are of similar size and the antago-nism escalates into a fight. This would never be the case in interactions between territorial males and females in nature (SL of territorial males: 80.9±0.5 mm [n=28]; random sample of 45 adult females: 63.5±0.4 mm). We measured terri-tory size (m2) of each male; territory borders were determined by the aggressive
behaviour that the owner showed towards other fish (con- and heterospecific). Af-ter completion of behavioural data collection, all males were caught either in gill-nets or by hook and line, both using SCUBA. Immediately thereafter males were measured (SL; to the nearest 0,1 mm) and photographed for colour analysis.
Mesocosm experiment
We built three outdoor concrete ponds of about 3·2·1 m3 (length·width·depth)
each at the Mwanza station of the Tanzania Fisheries Research Institute (TAFIRI), on the eastern shore of the Mwanza Gulf. After collection from Makobe Island, fish were kept in groups of approximately 40 individuals in larger ponds for three weeks. After this acclimatisation period we finclipped all fish, photographed the males and assembled three groups of six adult males and 20 adult females that were released in each pond. Twenty PVC tubes (length 15 cm, diameter 10 cm) in each pond provided shelter. For a period of three weeks, starting one week after releasing the fish, we collected the eggs of every brooding female (10, 11 and 7 clutches for groups 1, 2 and 3). To do so, ponds were seined once a week. Average clutch size was 48±5 eggs; all eggs were preserved in absolute ethanol. We used the number of matings, i.e. the number of clutches completely or partially sired by a male, as a measure of male reproductive success.
poly-merase (Bioline), 2.0mM MgCl2 and 1μl 10x Mg free reaction buffer (Bioline).
PCR programs were: Ppun5: two step protocol: 3 min at 94 °C, followed by five cycles of 94 °C for 30s, 62 °C for 30s and 72 °C for 30s followed by thirty cycles of 94 °C for 30s, 60 °C for 30s and 72 °C for 30s, followed by one cycle of 10 min of 72 °C. Ppun7: standard protocol: 3 min at 94 °C, followed by thirty cycles of 94 °C for 30s, 57 °C for 30s and 72 °C for 30s, followed by one cycle of 10 min of 72 °C. Ppun20 and Ppun21: touch down protocol: 3 min at 94 °C, followed by eight cy-cles of 94 °C for 30s and 60 °C dropping one degree every cycle (down to 53 °C) for 30s and 72 °C for 30s, followed by 25 cycles of 94 °C for 30s, 53 °C for 30s and 72 °C for 30s, followed by one cycle of 10 min of 72 °C.
The result of the PCR reaction was checked on a 1.5% agarose gel together with a 100 bp ladder. Amplification products were resolved on an ABI 377 se-quencer, processed in GeneScan 3.1 and scored in Genotyper 3.1 (Perkin-Elmer). Of 260 analysed offspring, 13 were discarded because of insufficient DNA. Exclu-sion probability was 99.8% and all males were excluded as sires except one for all remaining 247 larvae (in 28 clutches; 8.8±0.3 offspring per clutch).
Colour analysis
All males were photographed under standard conditions, placed in a perspex cu-vette with water and gently squeezed between a grey PVC sheet and the front window. We used an SLR camera with 100 mm lens and one flash on either side. All pictures were digitised. In Photoshop 6.0 (Adobe Systems Inc.) we adjusted white balance with the aid of a white patch that was attached to the front side of the cuvette. We analysed the colours of the fish body, excluding fins and eyes, in SigmaScan Pro 4.0 (SPSS Inc.). To calculate colour scores, we defined criteria to delimit the body area covered by ‘red’ and ‘yellow’ by a combination of hue and saturation (red: hue=0-26 plus 232-255, saturation 40%-97%; yellow: hue=27-45, saturation 40-97%) and subsequently calculated the area of the fish body that matched these criteria, yielding a percentage of body coverage. We similarly de-fined criteria for ‘blackness’, to quantify the body coverage of the black vertical bars and ventral aspects of the body (black: intensity=0-75).
Data analysis
Comparisons of groups and bivariate relationships were analysed using paired t-tests and Pearson correlations for normally distributed data, and Wilcoxon signed ranks tests (for dependent samples), Mann-Whitney-U tests (for independent samples) and Spearman correlations for non-normally distributed data (SPSS 10.0; SPSS Inc.). Means of normally distributed data are given with standard er-rors.
D I R E C T I O N A L S E X U A L S E L E C T I O N
total number of matings in each group and therefore analysed using binomial models as well. Female response to male characteristics in the field were recorded as counts and analysed using models with a Poisson distribution and a log link function. All generalised linear models were calculated in R (Ihaka & Gentleman 1996; http://www.r-project.org). Stepwise removal of nonsignificant variables from saturated models yielded minimal adequate models; significance was determined by F-tests examining the change in deviance following removal of each variable. We checked for over- and underdispersion and adjusted test statistics (Venables & Ripley 2002).
Results
Mate choice experiment in the laboratory
Females responded stronger to the courtship of males with high redscores (paired t-tests comparing average response to both males over all trials per female: Lateral Display t10=3.88, p=0.003; Quiver t10=3.08, p=0.012; Lead Swim t9=2.30,
p=0.047; Figure 2.1a). To test the relative importance of male coloration and size
female response (approach pe r male display) 0.0 0.2 0.4 0.6
response to male with highest redscore response to male with lowest redscore
Lead Swim * Lateral Display ** Quiver * high redscore low redscore 0 10 20 30 prop ort ion o f f ema les pr ef er rin g r edd er ma le ( % ) 0 25 50 75 100 0 10 20 30
redscore difference (% body coverage) 0 10 20 30
Lateral Display Quiver Lead Swim
b) a)
we analysed female preference using generalised linear models with a binomial distribution, with logit-transformed female response as a dependent variable and male coloration (red, yellow and black), standard length and weight as independ-ent variables. Only male redscore significantly influenced female response to male Lateral Display (estimate=4.25±0.79, F1,47=35.13, p<0.0001) and Lead Swim
(es-timate=3.44±1.05, F1,43=13.29, p<0.001). Female response to male Quiver was
also best explained by male redscore (estimate=4.98±1.11, F1,46=10.93, p<0.002)
but significantly decreased with male blackscore (estimate=-1.51±0.65, F1,46=
11.78, p<0.002). Besides the variation explained by covariates, there was no sig-nificant variation in response between females or between male pairs. Female re-sponsiveness did not increase with the redness means of males in a pair (Spear-man rank correlations: n=7, p>0.7), confirming that it was the difference in redscore between the two males rather than the absolute redscore of the preferred male that explained female response.
Males with high and low redscores did not differ in courtship behaviour, aggression or encounter rates with females (Wilcoxon Signed Ranks tests on me-dians of trials per male: n=7 male pairs, Z<1.19, p>0.24) and courtship intensity was not correlated with redscore (Spearman rank correlation of medians of trials per male: n=14 males; p>0.35 for each of the courtship behaviours). Female pref-erences were not correlated with behavioural diffpref-erences between the males (n=48 trials; p>0.45).
In order to define the threshold above which females start discriminating between male redscores, we carried out a generalised linear model analysis with the difference in redscore as single explanatory variable and the number of fe-males preferring either male as binomially distributed dependent variable. Figure 2.1b reports the predicted proportion of females preferring the reddest male in a pair, for any given difference in redscore between them (estimate [Lateral Dis-play]=0.12±0.036, F1,47=20.56, p=0.004; estimate [Quiver]=0.054±0.017, F1,43=
13.12, p=0.011, estimate [Lead Swim]=0.026±0.022, F1,39=1.42, p=0.29). The
curves become less steep as the escalation of courtship behaviours proceeds, and the curve for Lead Swim is not significant. This could be a sample size effect since not all courtship bouts proceeded to higher escalation levels, resulting in decreas-ing numbers of trials that met calculation criteria in the sequence Lateral Display-Quiver-Lead Swim; and decreasing numbers of observations within trials in the same sequence. For Lateral Display, calculation criteria were met in all 48 trials (Lateral Display frequency 31.7±4.1 per male per trial), for Quiver in 44 trials (23.9±3.6) and for Lead Swim in 40 trials (18.8±3.1). The curves for Lateral Dis-play and Quiver, both significant, indicate that females started discriminating as soon as there was a difference in redscore between the males.
Reproductive success in nature
D I R E C T I O N A L S E X U A L S E L E C T I O N
and female response to male courtship as dependent variable. Male redscore and courtship intensity significantly explained female response (estimate [redscore]= 0.046±0.016, F1,26=10.25, p=0.004; estimate [courtship intensity]=0.415±0.142,
F1,25=7.66, p=0.011; Figure 2.2).
These effects were independent: redder males did not court more intensely (Pearson correlation, n=27 [one male did not court], r=0.16, p=0.44). The red-scores of the males ranged from 0.3 to 28.7 % body coverage (15.4±1.5 %); a range similar to that among the males used in the laboratory experiment (3.3 to 34.3 %; 18.4±2.8 %).
There was a significant interaction between territory size and depth. Terri-tories were clustered in two depth categories, with 11 males at 4.2±0.1 m depth and 17 males at 6.3±0.1 m depth. Territories in shallow water were larger than those in deeper water (4.14±0.81 m2 and 2.03±0.36 m2; t
26=2.70, p=0.012).
Depth did not influence female choice, and redscore and courtship intensity did not correlate with depth or territory size. However, female choice for high red-score and courtship intensity also selected for males with territories that were large relative to others at the same depth (generalised linear model explaining female response by depth and territory size: estimate [territory size]= 0.137± 0.060, F1,25=4.74, p=0.039).
Mesocosm experiment
In two groups, three males (out of six) sired all genotyped offspring; in the third group, only two males (out of six) sired all offspring (Table 2.1). The majority of females had mated with more than one male: 19 of 28 broods had two or more fathers (68%; mean number of fathers per brood=1.8±0.1 across all broods). There was no correlation between the number of genotyped fry in a clutch and the number of fathers assigned (Spearman rank correlation: n=28 broods, rs=0.16, p=0.41), indicating that we did not significantly underestimate the
num-ber of sires due to incomplete genotyping.
redscore (% body coverage)
0 5 10 15 20 25 30 fe male respon se (p er 10 minu tes) 0.1 0.2 0.4 0.6 0.8 1
courtship intensity (Quiver per female)
0 1 2 3 4
To determine which male characteristics influenced mating success, we calculated a generalised linear model with binomial distribution and logit link function using blackscore, redscore, yellowscore, male size (standard length) and ‘group’ as inde-pendent variables and the number of matings as deinde-pendent variable. The slopes of the relationships between male traits and the number of matings did not differ between the groups (all F1,14<0.86, p>0.37). The minimal adequate model to
de-scribe mating success included blackscore only (estimate=2.19±0.76, F1,16=9.02,
p=0.008; Figure 2.3a, left panel). Including the effect of redscore did not improve the model (F1,15=0.32, p=0.58). When blackscore was excluded from the model,
there was a nearly significant trend for redder males to sire more offspring (esti-mate=1.76±0.91, F1,16=4.0, p=0.063; Figure 2.3a, right panel). Redscore and
blackscore were strongly correlated (Spearman rank correlation: rs=0.76,
p=0.0004).
The unmated males in the experiment had significantly lower blackscores than the mated males, and than the territorial males in our field study (Mann-Whitney U tests: compared to mated males in the experiment: n1=10, n2=8, Z=
2.67, p<0.01; compared to field males n1=10, n2=28, Z= 3.35, p<0.001; Figure
2.3b); whereas the mated males in the experiment did not differ from the territo-rial field males (n1=8, n2=28, Z= 0.50, p=0.64). This suggests that the males with
low blackscores in our experiments did not become territorial, which in turn im-plies that most unmated males possessed no territory. This makes the mesocosm data set different from our field and laboratory data sets which consisted entirely of territorial males. Because mesocosm males were photographed prior to the ex-periments, this also suggests that blackscore predicts future territoriality. We therefore repeated the analysis including only those males that had mated (n=8; 7 with high blackscores, one with low). In this subset, redscore emerged as the best, but again non-significant, predictor of mating success (estimate [redscore]= 1.38±0.81, F1,7=3.05, p=0.13; estimate [blackscore]=0.61±0.74, F1,7=0.70, p=
0.44). Redscore and blackscore were no longer correlated (rs=0.19, p=0.5). Thus,
this smaller data set resembled the territorial field males among which redscore and blackscore were not correlated either (rs=-0.15, p=0.44).
Table 2.1. Summary of the results of the mesocosm experiment.
group 1 2 3
number of mated males 3 3 2
number of clutches 10 11 7
number of genotyped fry 89 92 69
number of clutches with
one father (% of clutches) 3 (30) 4 (36) 2 (29)
two fathers (% of clutches) 4 (40) 6 (55) 5 (71)
three fathers (% of clutches) 3 (30) 1 (9) 0 (0)
D I R E C T I O N A L S E X U A L S E L E C T I O N
Discussion
Two different mate choice experiments and a field study all suggest that the red-ness of males is the most important criterion for female mate choice among terri-torial males in Pundamilia nyererei. In a behavioural laboratory experiment, male redness predicted female response to courtship displays. Probability analysis indi-cated that the proportion of females preferring the redder one of two males in-creased with the magnitude of the difference between them. The predicted pro-portion of redder-preferring females exceeded 50% as soon as there was a difference between the males, suggesting that even a very small difference in red-ness might be enough to elicit a preference, and associated variation in mating
blac
kscore (% body coverage)
0.01 0.1 1 10 high blackscore low blackscore field territorial n=28 matedn=8 unmatedn=10 mesocosm blackscore (% of group maximum) 0 20 40 60 80 100 ma ting s (% o f gro up total) 0 10 20 30 40 50 redscore (% of group maximum) 0 20 40 60 80 100 ma ting s (% o f gro up total) 0 10 20 30 40 50 b) a)
success on population level. However, our experiment included only two male pairs with redscore differences below 10% body coverage. We suspect that we would detect a female response threshold with more trials in this range. Male size and yellowscore did not influence female response; male blackscore significantly decreased female response to male Quiver but did not affect response to Lateral Display or Lead Swim.
The variation in redscores among males in the aquarium experiment re-sembled that among the field males. In nature however, females can use cues for mate choice decisions that were excluded in the experiment, such as chemical sig-nals, territory size and water depth. Our field data reveal that even when these cues are available, male redness is the best predictor of female mate choice, with territory size a secondary factor. Male courtship intensity also predicted female response, possibly due to an interaction between male and female behaviour, with males displaying more as females stay in their territory longer (Collins 1994). In the laboratory experiment, the male enclosures did not allow females to actually follow a male into his spawning pit, making such behavioural interactions less in-fluential. Results of the mesocosm mating experiment under semi-natural condi-tions are consistent with the field and laboratory results, but apparently highlight the importance of male territoriality for mating success. This may have implica-tions for the potential of intrasexual selection (Seehausen & Schluter 2004; Van Doorn et al. 2004). The high incidence of multiple paternity that we observed re-sembles estimates for Lake Malawi haplochromines (Kellogg et al. 1995; Parker & Kornfield 1996; Knight & Turner 2004). In the field we frequently observed males courting females that were already mouthbrooding, illustrating the poten-tial for multiple mating in nature.
pa-D I R E C T I O N A L S E X U A L S E L E C T I O N
ternal investment is confined to costs related to territoriality. We also never ob-served parasitic spawnings by non-territorial ‘sneaker’ males in the field, but spawning was observed only five times in 195 focal observations.
Unfortunately, the water in the ponds quickly became rather turbid and this made direct observations of male territoriality impossible. At the same time, high algal density hampers light transmission and selectively removes red light. This may provide an additional explanation for the finding that male reproduc-tive success was not significantly related to redscore.
The correlation between redscore and blackscore among mesocosm males disappeared once unmated males were excluded, suggesting that red coloration is not fully expressed until males establish territories, a plausible strategy if the ex-pression of coloration is costly. As a consequence, female choice would have been limited to only two or three males in each mesocosm group. With two or three territorial males in each pond, the territory density resembles the field situation in which territories typically occupy about three square metres.
We conclude that the conspicuous red coloration of male P. nyererei which is important in behavioural reproductive isolation between this species and its blue sister species (Seehausen & Van Alphen 1998), is subject to directional sexual se-lection by female mate choice within P. nyererei. This is consistent with a speciation scenario in which sexual selection through female choice has played an important role during the divergence of P. nyererei from P. pundamilia. Although we cannot currently prove it beyond doubt, all evidence suggests that P. pundamilia with me-tallic blue-grey males and females preferring them over red males (Seehausen & Van Alphen 1998), represents the ancestral condition (Seehausen et al. 1997a). P.
pundamilia has a wider and more continuous geographical distribution, whereas that of P. nyererei is nested within the latter; and blue-grey is the dominant male colour pattern in the genus Pundamilia in general (Seehausen & Van Alphen 1999).
Colour polymorphisms in male nuptial coloration, involving blue-red or blue-yellow sister species and intraspecific morphs are common in haplochromine cichlids (Seehausen et al. 1999c). Here we have shown that male nuptial coloration is also under directional intersexual selection within a species. Similar results were recently obtained in a cichlid population of Lake Malawi (Pauers et al. 2004). The next challenge is to explain why and when such sexual selection becomes disrup-tive. This will require investigations into proximate (e.g. sensory) and ultimate causes of individual variation in mating preferences (Jennions & Petrie 1997).
